Prematurity and the Transition at Birth

Why Birth Creates a Modeling Question

A preterm infant’s development does not begin at birth. By the time the infant is born, development has already occurred in utero; after birth, development continues under a different set of physiological conditions. A model that uses only time since birth can therefore treat infants with very different developmental histories as if they had the same starting point. A model that uses only developmental age can make the opposite simplification: it can treat an infant who has spent weeks outside the uterus as equivalent to one who has spent those weeks in utero.

The earlier material on gestational, postnatal, and postmenstrual age defines the age scales used here. In brief, gestational age describes development before birth, postnatal age describes elapsed time since birth, and postmenstrual age (PMA) combines gestational age at birth with postnatal age. The modeling question is not simply which age is “correct.” It is whether the process being modeled follows developmental age, time since birth, or both.

Definition: Let \(G_0\) be gestational age at birth and let \(t\) be postnatal age, expressed in compatible units. Then postmenstrual age at observation is \(P=G_0+t\). A PMA-based maturation model describes change by \(P\); a postnatal transition term describes a separate change with \(t\). These variables encode different aspects of developmental history and should not be used interchangeably.

PMA can be a useful developmental clock, but it does not record where development occurred. For example, two infants observed at 32 weeks PMA have the same PMA, while one was born at 28 weeks and has spent four weeks postnatally, and the other was just born at 32 weeks. A model using PMA alone predicts the same maturation for them. That prediction is an assumption, not a consequence of how PMA is calculated.

A Developmental Curve and a Birth-Transition Term

A common way to express a bounded maturation process is the Hill function. If \(P\) is PMA, \(P_{50}\) is the age at which the modeled fraction is one-half, and \(h>0\) controls the steepness, define

F(P)=\frac{P^h}{P^h+P_{50}^h}.

This function is between zero and one for positive \(P\). At \(P=P_{50}\), the numerator and the second denominator term are equal, so \(F(P_{50})=1/2\). The curve describes a developmental pattern indexed by PMA; by itself it does not represent a distinct effect of having been born.

One possible way to represent a postnatal transition is a separate multiplier:

T(t)=\exp\!\left[-A\exp(-kt)\right],\qquad A>0,\;k>0,\;t\geq 0.

This is a candidate model component, not a universal law of neonatal physiology. It begins at \(T(0)=\exp(-A)\), increases with postnatal age, and approaches 1 as \(t\) increases. A process described by both components might be represented by the normalized quantity \(M(P,t)=F(P)T(t)\). The interpretation is that \(F\) represents a PMA-related developmental fraction and \(T\) represents an additional postnatal transition in that same modeled quantity.

Model assumptions: Using \(M(P,t)=F(P)T(t)\) assumes that the PMA-related developmental pattern and the postnatal transition act multiplicatively on the chosen quantity; that \(T(t)\) has the same form and parameters across the individuals to whom it is applied; and that the chosen scale permits a product of the two factors. It also assumes the transition increases toward its reference level. If the transition is pathway-specific, differs by gestational age at birth, or has the opposite direction, this particular form may be unsuitable.

Worked Calculations

Worked Example: Maturation at Two PMA Values

Suppose a hypothetical pathway is described by a Hill function with \(P_{50}=32\) weeks and \(h=4\). Calculate its maturation fraction at 28 and 40 weeks PMA. These calculations concern the PMA component only; they do not establish that a birth-transition term is needed.

At 28 weeks,

F(28)=\frac{28^4}{28^4+32^4} =\frac{614{,}656}{614{,}656+1{,}048{,}576} =\frac{614{,}656}{1{,}663{,}232} \approx 0.3696.

At 40 weeks,

F(40)=\frac{40^4}{40^4+32^4} =\frac{2{,}560{,}000}{2{,}560{,}000+1{,}048{,}576} =\frac{2{,}560{,}000}{3{,}608{,}576} \approx 0.7094.

Under this hypothetical curve, the modeled fraction rises from about 0.3696 to 0.7094 as PMA increases from 28 to 40 weeks. These are dimensionless fractions, not direct clearance values. Turning them into a pharmacokinetic parameter requires a model linking that parameter to the fraction.

Worked Example: Same PMA, Different Time Since Birth

Consider two hypothetical infants observed at 32 weeks PMA. Infant A was born at 28 weeks and is observed at postnatal day 28. Infant B was born at 32 weeks and is observed at birth. Use the curve from the first example and a transition model with \(A=0.30\) and \(k=0.05\) per day. Calculate the predictions of the candidate model \(M(P,t)=F(P)T(t)\).

Both infants have \(P=32\) weeks, so the PMA component is \(F(32)=1/2=0.5000\). For Infant A, \(t=28\) days:

T(28)=\exp[-0.30\exp(-0.05\times28)] =\exp[-0.30\exp(-1.4)] \approx \exp(-0.0739791) \approx 0.9287.

Thus \(M(32,28)=0.5000\times0.9287\approx0.4643\). For Infant B, \(t=0\), so \(T(0)=\exp(-0.30)\approx0.7408\), and \(M(32,0)=0.5000\times0.7408\approx0.3704\). The difference arises entirely from the postnatal term: the PMA-only model would predict 0.5000 for both infants. Whether a difference of this kind is supported by data and physiology must be assessed rather than assumed.

Worked Example: Checking the Transition’s Time Course

For the same hypothetical transition parameters, evaluate \(T(t)\) at birth, 70 days, and 84 days. This illustrates the transition term alone, not a complete maturation prediction.

At birth, \(T(0)=\exp(-0.30)\approx0.7408\). At 70 days, \(0.05\times70=3.5\), and \(\exp(-3.5)\approx0.0301974\). Therefore,

T(70)=\exp[-0.30\times0.0301974] =\exp(-0.00905922) \approx0.9910.

At 84 days, \(0.05\times84=4.2\), and \(\exp(-4.2)\approx0.0149956\). Hence

T(84)=\exp[-0.30\times0.0149956] =\exp(-0.00449868) \approx0.9955.

The factor approaches 1 over time but does not equal 1 at a finite time. The numbers illustrate the parameterization; they do not show that a biological process must follow this time course.

Why the Transition Form Has These Properties

The boundary values and direction of change follow directly from the definition. At \(t=0\), substituting zero gives \(T(0)=\exp[-A\exp(0)]=\exp(-A)\). As \(t\) tends to infinity, \(\exp(-kt)\) tends to zero, so \(T(t)\) tends to \(\exp(0)=1\). Differentiating with respect to postnatal time gives

\frac{dT(t)}{dt} =T(t)\,A k\exp(-kt)>0.

Thus the function rises monotonically for \(A>0\), \(k>0\), and \(t\geq0\). The initial value is controlled by \(A\); the rate of approach is controlled by \(k\), which has units of inverse time. For example, if \(t\) is in days, \(k\) is in inverse days. These interpretations depend on the stated form and should not be generalized to every model of transition at birth.

Choosing and Checking an Age Structure

The essential modeling decision is which aspect of history is expected to govern the parameter of interest. A PMA-only curve is appropriate when the same developmental age is expected to imply the same modeled fraction, regardless of whether the time was spent in utero or after birth. A postnatal-age model instead describes change from birth and may be useful for a process tied to adaptation after delivery. A model with both terms is a more specific hypothesis: it says that developmental age and postnatal transition contribute separately, in the stated functional form.

Model description	What it treats as relevant	Main assumption or limitation
PMA-only maturation	Developmental age, including time before birth	Equal PMA implies equal modeled maturation, regardless of birth timing
Postnatal-age curve	Elapsed time since birth	Does not, by itself, encode how much development occurred before birth
PMA curve with transition factor	Developmental age and a separate postnatal process	Requires evidence for the extra term and for its functional form and shared parameters

Diagnostics should focus on whether the model’s assumptions leave systematic patterns. Examine whether residuals or prediction errors differ by postnatal age among infants with comparable PMA, or differ by PMA among infants with comparable postnatal age. Plot observations and predictions against both age scales, and mark gestational age at birth or birth groups where the data allow. Consistent underprediction early after birth followed by improvement with postnatal time could signal a missing transition component; a pattern confined to a metabolic pathway may instead argue against imposing one shared transition across pathways.

A transition term can also fail because the study contains too little information to distinguish it from the PMA curve. In particular, if postnatal age and PMA move together over the observed records, their contributions may be difficult to separate. Implausible estimates, large uncertainty, strong dependence on starting values, or little improvement in predictive checks are warnings against treating a more elaborate model as established. A transition should be retained because it is supported by the data and scientific rationale, not simply because prematurity is clinically important.

Common Pitfalls and Limitations

Using postnatal age as if it were developmental age. An infant born preterm has already developed before birth. Time since birth alone does not encode that history.
Assuming PMA fully captures history. PMA is a useful age scale, but a PMA-only model assumes that birth timing adds no further influence on the modeled process.
Calling a transition multiplier a universal maturity correction. \(T(t)\) is a proposed model component. It is not automatically applicable to every enzyme, transporter, or clearance pathway.
Confusing a model prediction with a causal effect of birth. A fitted postnatal-age term describes a statistical pattern under model assumptions. It does not by itself establish that birth caused the pattern.

The multiplicative form \(F(P)T(t)\) constrains the modeled quantity to a particular structure. It cannot represent every interaction, a transition that depends on gestational age at birth, a transition that decreases after birth, or different transitions for distinct pathways without further terms. Those extensions require sufficient data and a clear physiological interpretation. In addition, the factors are normalized fractions in the examples; they should not be reported as clearance, volume, or another parameter with physical units unless a parameter model supplies the appropriate scale and units.

Key takeaway: PMA records developmental age, while postnatal age records time since birth. For preterm infants these histories are not interchangeable. A model may use PMA alone, postnatal age alone, or both, but each choice makes a distinct, testable assumption about the process being modeled.

Check Your Understanding

Answer each question, then compare your reasoning with the solutions.

An infant is born at 30 weeks gestational age and is observed 3 weeks later. What is the infant’s PMA, and what does postnatal age measure?
Under a PMA-only model, what is predicted for two infants observed at the same PMA but with different postnatal ages? State the assumption this prediction makes.
For \(T(t)=\exp[-0.20\exp(-0.10t)]\), calculate \(T(0)\) to four significant figures.
Name one pattern in model predictions or residuals that could motivate checking for a postnatal transition term.

Solutions. (1) The PMA is \(30+3=33\) weeks; postnatal age is the elapsed time since birth, here 3 weeks. (2) The model predicts equal maturation at equal PMA. It assumes birth timing and time since birth do not add further information once PMA is known. (3) \(T(0)=\exp[-0.20\exp(0)]=\exp(-0.20)\approx0.8187\), a dimensionless value. (4) For example, systematic underprediction at short postnatal ages among infants with similar PMA could motivate checking the transition hypothesis, while not proving that it is correct.